Imagine a billionaire kidnapped a random person and said: Here is a camera, banana, bicycle, sundress, and porcupine; you have one day to, by yourself, make a funny video using them all. I’ll upload it to YouTube, and if it gets “enough” views in the first day, I’ll give you $10B. If not, I’ll kill you and your closest twenty relatives. No, I’m not going to tell you my “enough” threshold. Go.
I'm much more optimistic about a return of sortition than you are. One of Classical Athens' dirty little-known secrets (and that of less cool poleis) is that it relied on sortition for much of its history, particularly after the Peloponnesian War made the shortcomings of democratic populism evident to all. Sortition could be useful for all sorts of appointments, particularly if you (like the Athenians did) restrict the pool of candidates to those without a criminal history, property-owners, etc. Partitocracies across the West would lose their whole reason for existing.
Sortition is used in citizens' assemblies. They address your concerns by giving participants time and information to become more informed than general public opinion. Then they must find agreement on answers to their topic/issue/problem.
There are North American organisations like Healthy Democracy, American Public Trust and MASSLBP working in this space. The Democracy R&D Network is the global network for 'sortition' and 'deliberation'.
I work in this space and would be happen to answer anyone's questions.
Example 1 is a DISHONEST hijacking of an idea. "Capitalism incentive" applies to a population (as opposed to a specific individual) over an extended period (as opposed to one day). the whole argument in example 1 is distasteful.
My first deep impression is that I fail to see how this constitutes a case about the limitations of incentives. Initially, it seems, more simply, a way to say random selection is inferior to goal-driven self-selection, which is uncontroversial. Upon further inspection, I realized the root cause of my suspicion - self-selection is, itself, a product of incentives.
Basically, the megastructure of all value creation is 'incentives all the way down'. Thus, the notion that 'incentives alone' < 'incentives + competition' appears illuminating only when one does not acknowledge that competition is a pressure pot of drive and incentives.
Concerning a general model for imagining instances of the basic scenario, let us consider a system with two categories of features, Attraction and Exploitation, to which we can apply two variants of incentives - Narrow (NI) and Open (OI). NI corresponds to selection, and OI to competition (we'll see how below).
Consider an organization seeking talented individuals to achieve a creative goal. The Attraction features include publicity, screening, and other recruitment processes. The Exploitation features include the internal mechanisms within the organization that enable the attracted talent to achieve the creative goals (actually, this could be a 3-part system with onboarding and orientation being in the middle portion called Situation).
Applying NI/OI to Attraction: if we select the candidates randomly - applying NI, we restrict the interplay of incentives, thus increasing the single-dimensional incentive infinitely does not improve the overall outcome beyond chance. We can make a few interesting inferences here, but let us assume the effective incentive value is 0. On the other hand, if we apply OI, the emerging competition entails higher dimensions of interplaying incentives and so more skill results in higher productivity. Consequently, the most capable candidates (at gaming the incentives) get in.
NI/OI to Exploitation: If one standardizes the incentives for producing creative outputs, such as paying a fixed sum to all the hires for an expected result, adjusting the fixed sum may have no effect on the probability of achieving a given result (although an OI-Attraction team does better on average). However, by using the same resources to incentivize creative output by paying variable sums to the hires based on their score on some chosen metric of success, the chances that extraordinarily good outcomes are achieved further increase.
In essence, the optimal combination is OI-Attraction × OI-Exploitation!
The possible scenario combinations and their illustrations are:
1. OI-A × NI-E => Creative media agencies
2. OI-A × OI-E => Social media platform e.g. YouTube
3. NI-A × OI-E => "I have chosen you; I will give you a reward proportionate to your performance on a given metric"
4. NI-A × NI-E => "I have chosen you; I will give you an arbitrary, potentially infinite reward for a stated outcome"
Something like the above looks like the universal picture. We can imply that the first example in the original letter is an instance of combination 4 and is demonstrably set up for failure. Yet, no incentives were harmed in the making of this documentary.
Q.E.D
In short, this could have been an essay about why incentives are intricately critical, with the alternate title When Incentives Are Not Enough, Create More.
It's very difficult for the elites to accept that ten thousand average people make far better economic and political decisions than ten highly-qualified experts, even discounting that the ten experts will always say what they are paid to say. Also, very few if any political decisions take rocket surgeons to predict the outcome. The polis know but go ahead anyway, as the only outcome that interests them is re-election. As far as I can see, prediction markets are simply a respectable method of gambling and influence the future by not one act of sexual intercourse. Moral, to make the world a better place, keep the self-selected aka activists as far from public decision-making as you can.
"incentives are not enough" - an insight obvious to anyone who has tried to teach a dog, a child, or a neural network! Learning requires time and data, and a strong incentive is often actually worse than a weak one because it inhibits exploration and increases overfitting / Goodharting risks. (hence e.g. the institution of tenure).
So "capitalism" doesn't feel right as a description. You're talking about a theory of economic learning, of cultural knowledge, of human & organizational capital accumulation. Yes free markets can speed up such learning. But if we look at the most valuable human & organizational capital in the world, much of it was built through intense ongoing state investment (e.g. decades of Taiwanese industrial policy created TSMC; decades of communist investment in science & math education is still responsible for a substantial fraction of the world's top STEM talent).
Managing a learning process is a hard and multidimensional optimization problem, and "give more freedom" is probably not much more useful than "crank up the the incentives" as a general guidance.
Maybe I'm over-fixating on this one point, but the futarchy idea seems to fail without additional sophistication. It also requires tradeoff rates, conditionals, and diminishing marginal utility for Values. Without them there's the potential to get stuck in local maxima where the market maximizes investment into the easiest Values even though it comes at a detriment to other Values. Voters may also only want A to happen if B happens, since either alone may not be in their interest.
If I value three things – fertility rate, GDP, and my purchasing power – I may have a tradeoff rate where I'm willing to sacrifice X GDP per unit increase in TFR. However, growing GDP is probably easier for Western nations than raising the fertility rate, so the policy betting market for GDP growth will receive far more attention than the fertility market. If you're a policy person interested in making money on these market, there's a headwind to go after the tougher problems and you may just opt for easier, GDP-benefiting policies.
Also, getting women to spend less time on family and more on economic activities is a plausible method of growing GDP. Unless the voters can specify that they don't want A to happen to the detriment of B, policy markets for GDP growth could undermine the efficacy of policy markets for fertility gain.
Voters may also only care about GDP increases when they also increase the median purchasing power. That way policies where GDP growth is mostly captured by a tail of the population won't be passed. This is simply a loss of information when distilling a Value to a metric, so concordant metrics are required to capture the underlying Value.
Without diminishing marginal utility for Values, policy will become dominated by pro-GDP. Without a penalty function in place, policy could happily spite more difficult-to-implement Values for ones easier to maximize. Without conditional requirements or bundling of Value metrics, policies could successfully increase the metric without realizing the underlying Value.
I enthusiastically await the day a politician presents a PowerPoint covering their model of Values with MU curves, penalization strategies, and piecewise functions at a town hall.
Indemnities aren't super common. If you're with a big employer (you don't need to pass underwriting) you can get maybe $25k or so in most cases. To get that money you need to have cancer or something terrible. If you've got cancer, $25k is basically nothing compared to what your medical bills are going to be. And you likely have health insurance as well.
a bigger idea (partially implied) is allowing arbitrage on other persons lives and health.
Suppose can help improve one's English or another language. huge utility. but my value from pushing this on someone is far from his eventual utility. same with health, moving mutual funds to index etc.
You're right. It's important not only to give incentives, but also to allow many agents to compete for them, so that the best-performing agents do the job and get the incentives.
A better term for this would be "competition," not capitalism. Capitalism often involves anti-competitive factors, e.g. monopolies and near-monopolies often arise in capitalism. What you want is to promote a high degree of competition.
It would be interesting to see the state of the various predictions markets that exist: How much $$ is transacting, how they can be broadened in scope, etc. It always struck me as a concept that could have broad utility. For example I've known people that I would bet heavily on ending up in jail, and I would have won those bets. There are academic papers I would bet are wrong. And so on.
I guess I don't really think you're using incentive in the right way, or broadly enough. In your billionaire example, the billionaire randomly selects a person to probably murder, but the billionaire is not under any incentive to produce the given outcome. Likewise, in your examples where things are working well, it is because everyone has incentives to achieve the given outcome ie service utopia. I think many failures come down to someone in the process having bad incentives. In real life, this would involve a bureaucrat having incentives to win local status games instead of providing quality services. Your example of futarchy works because speculators are incentivized to give high quality information on important outcomes. It would still break if voters were not incentivized to pay attention to that information or elected officials were not incentivized to act on that information. Overall, while I think your examples are true, I think they are more cases of insufficient incentives or bad incentives for all players, rather than needing a separate concept like freedom
I'm much more optimistic about a return of sortition than you are. One of Classical Athens' dirty little-known secrets (and that of less cool poleis) is that it relied on sortition for much of its history, particularly after the Peloponnesian War made the shortcomings of democratic populism evident to all. Sortition could be useful for all sorts of appointments, particularly if you (like the Athenians did) restrict the pool of candidates to those without a criminal history, property-owners, etc. Partitocracies across the West would lose their whole reason for existing.
Sortition is used in citizens' assemblies. They address your concerns by giving participants time and information to become more informed than general public opinion. Then they must find agreement on answers to their topic/issue/problem.
There are North American organisations like Healthy Democracy, American Public Trust and MASSLBP working in this space. The Democracy R&D Network is the global network for 'sortition' and 'deliberation'.
I work in this space and would be happen to answer anyone's questions.
Example 1 is a DISHONEST hijacking of an idea. "Capitalism incentive" applies to a population (as opposed to a specific individual) over an extended period (as opposed to one day). the whole argument in example 1 is distasteful.
Are Incentives Not Enough?
My first deep impression is that I fail to see how this constitutes a case about the limitations of incentives. Initially, it seems, more simply, a way to say random selection is inferior to goal-driven self-selection, which is uncontroversial. Upon further inspection, I realized the root cause of my suspicion - self-selection is, itself, a product of incentives.
Basically, the megastructure of all value creation is 'incentives all the way down'. Thus, the notion that 'incentives alone' < 'incentives + competition' appears illuminating only when one does not acknowledge that competition is a pressure pot of drive and incentives.
Concerning a general model for imagining instances of the basic scenario, let us consider a system with two categories of features, Attraction and Exploitation, to which we can apply two variants of incentives - Narrow (NI) and Open (OI). NI corresponds to selection, and OI to competition (we'll see how below).
Consider an organization seeking talented individuals to achieve a creative goal. The Attraction features include publicity, screening, and other recruitment processes. The Exploitation features include the internal mechanisms within the organization that enable the attracted talent to achieve the creative goals (actually, this could be a 3-part system with onboarding and orientation being in the middle portion called Situation).
Applying NI/OI to Attraction: if we select the candidates randomly - applying NI, we restrict the interplay of incentives, thus increasing the single-dimensional incentive infinitely does not improve the overall outcome beyond chance. We can make a few interesting inferences here, but let us assume the effective incentive value is 0. On the other hand, if we apply OI, the emerging competition entails higher dimensions of interplaying incentives and so more skill results in higher productivity. Consequently, the most capable candidates (at gaming the incentives) get in.
NI/OI to Exploitation: If one standardizes the incentives for producing creative outputs, such as paying a fixed sum to all the hires for an expected result, adjusting the fixed sum may have no effect on the probability of achieving a given result (although an OI-Attraction team does better on average). However, by using the same resources to incentivize creative output by paying variable sums to the hires based on their score on some chosen metric of success, the chances that extraordinarily good outcomes are achieved further increase.
In essence, the optimal combination is OI-Attraction × OI-Exploitation!
The possible scenario combinations and their illustrations are:
1. OI-A × NI-E => Creative media agencies
2. OI-A × OI-E => Social media platform e.g. YouTube
3. NI-A × OI-E => "I have chosen you; I will give you a reward proportionate to your performance on a given metric"
4. NI-A × NI-E => "I have chosen you; I will give you an arbitrary, potentially infinite reward for a stated outcome"
Something like the above looks like the universal picture. We can imply that the first example in the original letter is an instance of combination 4 and is demonstrably set up for failure. Yet, no incentives were harmed in the making of this documentary.
Q.E.D
In short, this could have been an essay about why incentives are intricately critical, with the alternate title When Incentives Are Not Enough, Create More.
It's very difficult for the elites to accept that ten thousand average people make far better economic and political decisions than ten highly-qualified experts, even discounting that the ten experts will always say what they are paid to say. Also, very few if any political decisions take rocket surgeons to predict the outcome. The polis know but go ahead anyway, as the only outcome that interests them is re-election. As far as I can see, prediction markets are simply a respectable method of gambling and influence the future by not one act of sexual intercourse. Moral, to make the world a better place, keep the self-selected aka activists as far from public decision-making as you can.
"incentives are not enough" - an insight obvious to anyone who has tried to teach a dog, a child, or a neural network! Learning requires time and data, and a strong incentive is often actually worse than a weak one because it inhibits exploration and increases overfitting / Goodharting risks. (hence e.g. the institution of tenure).
So "capitalism" doesn't feel right as a description. You're talking about a theory of economic learning, of cultural knowledge, of human & organizational capital accumulation. Yes free markets can speed up such learning. But if we look at the most valuable human & organizational capital in the world, much of it was built through intense ongoing state investment (e.g. decades of Taiwanese industrial policy created TSMC; decades of communist investment in science & math education is still responsible for a substantial fraction of the world's top STEM talent).
Managing a learning process is a hard and multidimensional optimization problem, and "give more freedom" is probably not much more useful than "crank up the the incentives" as a general guidance.
Maybe I'm over-fixating on this one point, but the futarchy idea seems to fail without additional sophistication. It also requires tradeoff rates, conditionals, and diminishing marginal utility for Values. Without them there's the potential to get stuck in local maxima where the market maximizes investment into the easiest Values even though it comes at a detriment to other Values. Voters may also only want A to happen if B happens, since either alone may not be in their interest.
If I value three things – fertility rate, GDP, and my purchasing power – I may have a tradeoff rate where I'm willing to sacrifice X GDP per unit increase in TFR. However, growing GDP is probably easier for Western nations than raising the fertility rate, so the policy betting market for GDP growth will receive far more attention than the fertility market. If you're a policy person interested in making money on these market, there's a headwind to go after the tougher problems and you may just opt for easier, GDP-benefiting policies.
Also, getting women to spend less time on family and more on economic activities is a plausible method of growing GDP. Unless the voters can specify that they don't want A to happen to the detriment of B, policy markets for GDP growth could undermine the efficacy of policy markets for fertility gain.
Voters may also only care about GDP increases when they also increase the median purchasing power. That way policies where GDP growth is mostly captured by a tail of the population won't be passed. This is simply a loss of information when distilling a Value to a metric, so concordant metrics are required to capture the underlying Value.
Without diminishing marginal utility for Values, policy will become dominated by pro-GDP. Without a penalty function in place, policy could happily spite more difficult-to-implement Values for ones easier to maximize. Without conditional requirements or bundling of Value metrics, policies could successfully increase the metric without realizing the underlying Value.
I enthusiastically await the day a politician presents a PowerPoint covering their model of Values with MU curves, penalization strategies, and piecewise functions at a town hall.
Indemnities aren't super common. If you're with a big employer (you don't need to pass underwriting) you can get maybe $25k or so in most cases. To get that money you need to have cancer or something terrible. If you've got cancer, $25k is basically nothing compared to what your medical bills are going to be. And you likely have health insurance as well.
a bigger idea (partially implied) is allowing arbitrage on other persons lives and health.
Suppose can help improve one's English or another language. huge utility. but my value from pushing this on someone is far from his eventual utility. same with health, moving mutual funds to index etc.
if we can solve this!
You're right. It's important not only to give incentives, but also to allow many agents to compete for them, so that the best-performing agents do the job and get the incentives.
A better term for this would be "competition," not capitalism. Capitalism often involves anti-competitive factors, e.g. monopolies and near-monopolies often arise in capitalism. What you want is to promote a high degree of competition.
It would be interesting to see the state of the various predictions markets that exist: How much $$ is transacting, how they can be broadened in scope, etc. It always struck me as a concept that could have broad utility. For example I've known people that I would bet heavily on ending up in jail, and I would have won those bets. There are academic papers I would bet are wrong. And so on.
I guess I don't really think you're using incentive in the right way, or broadly enough. In your billionaire example, the billionaire randomly selects a person to probably murder, but the billionaire is not under any incentive to produce the given outcome. Likewise, in your examples where things are working well, it is because everyone has incentives to achieve the given outcome ie service utopia. I think many failures come down to someone in the process having bad incentives. In real life, this would involve a bureaucrat having incentives to win local status games instead of providing quality services. Your example of futarchy works because speculators are incentivized to give high quality information on important outcomes. It would still break if voters were not incentivized to pay attention to that information or elected officials were not incentivized to act on that information. Overall, while I think your examples are true, I think they are more cases of insufficient incentives or bad incentives for all players, rather than needing a separate concept like freedom
Sortision! I love it, but only one jury are chosen among an already aristocratic (meritocratic is a particular case) and experienced group.
https://forum.effectivealtruism.org/posts/PyqPr4z76Z8xGZL22/sortition